I have a binary classification problem with a tabular dataset that has almost 2 million rows. I have a special constraint is that the output of the model should be explainable. Apart from simple models like logistic regression, Random Forest,etc, I don’t know if there is a way to use a NN and to be able to give a consistent explanation for the output of the model.
I’ve already used SHAP and LIME but I’d like to know if you have other ideas in mind.
Thanks Zachary. I do use this for feature selection but I want something that would give me more insight into the output than just feature importance… But yes, I do have in mind (If I fail to find some other methods, that would be the way to go)
We have just started exploring the What If Tool, because we are similarly interested in exploring the explainability of various types of models. We have just started using it, but it seems promising.